[pve-devel] [PATCH proxmox/proxmox-openid] fix #5076: Added extra audience verification checks.
Dominik Csapak
d.csapak at proxmox.com
Mon Feb 17 09:54:12 CET 2025
On 2/6/25 13:01, Alexander Abraham wrote:
> Two things were added to the proxmox-openid crate to fix
> bug #5076: i) the function to require strict audience checking
> was called and ii) an extra verifier function was added to check
> if the configured audiences match the receieved audiences.
Hi,
first, it would be nice if the three relevant patches (proxmox/access-control/manager) would get a
combined cover-letter. that way it's easier to see that the patches
belong together.
aside from that, it would also be good if the commit message contain
a 'why'. The 'what' and 'how' should (most often) be self-evident from
the diff, but the why isn't most of the time.
E.g. a short sentence like: We want to verify additional audiences because ...
makes it much easier to reason about the intentions later on.
a few smaller comments inline
>
> Signed-off-by: Alexander Abraham <a.abraham at proxmox.com>
> ---
> proxmox-openid/src/lib.rs | 29 +++++++++++++++++++----------
> 1 file changed, 19 insertions(+), 10 deletions(-)
>
> diff --git a/proxmox-openid/src/lib.rs b/proxmox-openid/src/lib.rs
> index fe65fded..396f55cd 100644
> --- a/proxmox-openid/src/lib.rs
> +++ b/proxmox-openid/src/lib.rs
> @@ -1,10 +1,9 @@
> #![cfg_attr(docsrs, feature(doc_cfg, doc_auto_cfg))]
>
> -use std::path::Path;
> -
> use anyhow::{format_err, Error};
> use serde::{Deserialize, Serialize};
> use serde_json::Value;
> +use std::path::Path;
these two hunks seem unrelated (and wrong), please leave the
'std' imports seperate
>
> mod http_client;
> pub use http_client::http_client;
> @@ -53,6 +52,8 @@ pub struct OpenIdConfig {
> pub prompt: Option<String>,
> #[serde(skip_serializing_if = "Option::is_none")]
> pub acr_values: Option<Vec<String>>,
> + #[serde(skip_serializing_if = "Option::is_none")]
> + pub aud: Option<Vec<String>>,
> }
>
> pub struct OpenIdAuthenticator {
> @@ -204,21 +205,32 @@ impl OpenIdAuthenticator {
> .set_pkce_verifier(private_auth_state.pkce_verifier())
> .request(http_client)
> .map_err(|err| format_err!("Failed to contact token endpoint: {}", err))?;
> -
any special reason why you remove the whitespace here?
> - let id_token_verifier: CoreIdTokenVerifier = self.client.id_token_verifier();
> let id_token_claims: &CoreIdTokenClaims = token_response
> .extra_fields()
> .id_token()
> .expect("Server did not return an ID token")
> - .claims(&id_token_verifier, &private_auth_state.nonce)
> + .claims(
> + &((self.client.id_token_verifier() as CoreIdTokenVerifier)
is this cast here really necessary? AFAICS it shouldn't ?
> + .require_audience_match(true)
> + .set_other_audience_verifier_fn(|aud| {
> + let curr_aud: &String = &**aud;
clippy warns here:
deref which would be done by auto-deref
so you can just write:
let curr_aud: &String = aud;
> + if &self.config.client_id == curr_aud {
> + true
> + } else {
> + match self.config.aud.as_ref() {
> + Some(confd_auds) => confd_auds.contains(curr_aud),
> + None => false,
> + }
> + }
> + })),
> + &private_auth_state.nonce,
> + )
> .map_err(|err| format_err!("Failed to verify ID token: {}", err))?;
> -
why the white space removal here too?
> let userinfo_claims: GenericUserInfoClaims = self
> .client
> .user_info(token_response.access_token().to_owned(), None)?
> .request(http_client)
> .map_err(|err| format_err!("Failed to contact userinfo endpoint: {}", err))?;
> -
and here
> Ok((id_token_claims.clone(), userinfo_claims))
> }
>
> @@ -230,9 +242,7 @@ impl OpenIdAuthenticator {
> ) -> Result<Value, Error> {
> let (id_token_claims, userinfo_claims) =
> self.verify_authorization_code(code, private_auth_state)?;
> -
> let mut data = serde_json::to_value(id_token_claims)?;
> -
and here
> let data2 = serde_json::to_value(userinfo_claims)?;
>
> if let Some(map) = data2.as_object() {
> @@ -243,7 +253,6 @@ impl OpenIdAuthenticator {
> data[key] = value.clone();
> }
> }
> -
and here
IMO, white space cleanup can be fine, but please as a separate (upfront) patch,
so it does not pollute the actual patch
that said, in this case, I'd just leave the empty lines in place
> Ok(data)
> }
> }
More information about the pve-devel
mailing list